Summary A de Havilland DHC-8-100 (Dash 8) aircraft (registration C-GZKH, serial number117) operated by Provincial Airlines Limited was a passenger revenue flight from St. John's to Deer Lake, Newfoundland and Labrador, with 36passengers and 3crew on board. During the climb-out from St. John's, the indicated airspeed gradually decreased to the point that the aircraft entered an aerodynamic stall. The aircraft descended rapidly, out of control, losing 4200feet before recovery was effected approximately 40seconds later. The incident occurred during daylight hours in instrument meteorological conditions. There were no injuries and the aircraft was not damaged. Ce rapport est galement disponible en franais. Other Factual Information Weather Before departure, the flight crew checked the en route weather. The graphic area forecasts (GFA) for icing, turbulence, and freezing level, issued at 0911 Newfoundland daylight time1 and valid for 1530 on 27May2005, depicted the freezing level at 10000feet. The comments section of the GFA indicated the possibility of nil to light icing during the climb above the freezing level. The GFA forecast light to nil turbulence in the area of flight. There were no pilot reports or significant meteorological reports (SIGMETS) issued advising of icing conditions or turbulence on the planned route. Incident The aircraft took off uneventfully from St. John's at 1739. Aircraft weight at take-off was calculated to be 34096 pounds, about 400pounds below the maximum take-off weight. After take-off, the aircraft entered a right climbing turn. During the turn, the captain (pilot flying) engaged the autopilot. Flight data recorder (FDR) data show that, when the autopilot was engaged, the indicated airspeed (KIAS) was 162and the rate of climb was 1190feet per minute (fpm). After autopilot engagement, the rate of climb remained constant, while the IAS fluctuated slowly between 160and 170KIAS. During the climb, the captain's primary task was to monitor the performance of the aircraft as it flew under the control of the autopilot and to make the necessary adjustments to engine power settings during the climb. After the aircraft was established in the climb on the outbound heading, the first officer commenced the paperwork necessary for the departure report. FDR data show that there was little to no turbulence throughout the climb. FDR data also show that the outside air temperature was 5C at 7000feet; the air temperature decreased to below the freezing level at about 11000feet. At around 7000feet, the crew selected the engine anti-ice systems ON and the ignition switches to manual; however, they elected to leave the pneumatic de-ice system OFF. The aircraft was not equipped with an electronic ice detector, so crews must detect ice visually. Ice forms primarily on the wing leading edges and on an ice probe located in front of the cockpit window. The crew was looking for ice accumulation during the climb above 7000feet, but none was detected. Through about 8000 feet, the airspeed started a gradual decrease from 170KIAS over a period of five minutes. During this time, the vertical speed continued at a constant 1190fpmup. The gradual decrease in airspeed was detected when the first officer looked up from his paperwork, noted the decreased airspeed, and advised the captain. The captain then rotated the pitch control wheel on the flight guidance controller toward nose down (see Figure1) to increase the airspeed. While attempting the adjustment, the captain saw the aircraft's stick shaker activate, causing the autopilot to disengage. This occurred at 14800feet above sea level, at 104KIAS. The captain then began to manually fly the aircraft. Within a second of autopilot disengagement, the aircraft began to roll right and pitch down (see AppendixA - Flight Data Recorder Plot and AppendixB - Flight Data Recorder Plot- Engines). Immediately after the aircraft began to roll, it was noticed that there was ice on the left engine inlet. The roll angle increased to 64, the pitch angle went from 15nose up to 5nose up, and the aircraft vertical acceleration dropped to approximately 0.5g. The aircraft pitch then increased to 30nose up briefly before decreasing to 40nose down. These conditions are indications that the aircraft wing had fully stalled. However, the captain interpreted the indications as severe turbulence. The FDR data show that the aircraft underwent three distinct stalls during the loss-of-control event, with the third stall being the most severe. The data show that the control column position cycled rapidly back and forth as the stall developed, but was moved generally aft, remaining aft during all three stalls. There was significant aileron and rudder pedal movement during the event, but these controls were ineffective in regaining control and were in response to the aircraft's movement, rather than the cause of it. The data indicated that aircraft control was regained when the control column was moved forward. An aircraft's stall airspeed increases as a result of ice contamination. Also, stall warning devices may not be accurate when ice is present and may not provide a normal warning of stall. The aircraft's stick shaker will normally activate six to nine per cent above the stall speed, providing ample opportunity for the pilot to initiate stall recovery procedures. According to the flight manual, the stall speed for the aircraft's configuration and weight was 94.5KIAS. However, the aircraft stalled very shortly after stick shaker activation, at about 103KIAS. During the loss of control, there were heavy control column forces and severe buffeting. The aircraft descended rapidly, losing 4200feet before recovery was effected. Minimum IAS recorded by the FDR during the loss of control was 0KIAS. The actual minimum airspeed would have been higher, because airflow to the pitot tube was most likely disrupted due to excessively high angles of attack and side-slip. The maximum airspeed was 210KIAS; this was recorded during the stall recovery. The gload peaked at 2.24g during the recovery. The standard technique for a stall recovery is to immediately and simultaneously advance the control column to reduce the angle of attack, apply maximum power, and then level the wings when the aircraft has exited the stall. An ice-induced stall requires a recovery technique in which the control column is moved forward aggressively (altitude permitting) to reduce the angle of attack and trade altitude for airspeed. FDR data show that, after the stall, power remained unchanged and the control column moved aft of its pre-stall position for about 22seconds. The aircraft exited the stall when the control column was subsequently moved forward. Immediately after recovery, the crew observed that ice was building up rapidly on the aircraft fuselage. Airframe de-ice equipment was then selectedON. The pneumatic boots functioned when selected, and were effective in removing the ice. The crew requested a lower altitude to remain clear of icing conditions and continued to Deer Lake. After landing, the pilot reported a severe turbulence encounter to company personnel. A heavy turbulence check was subsequently carried out, and no damage was found. Company Information Provincial Airlines Limited had an approved Air Operator Certificate (AOC) to operate two Dash 8 aircraft under Section705 of the Canadian Aviation Regulations (CARs). The company had been operating the aircraft since January2004. The captain held a valid airline transport pilot licence (ATPL) and was certified and qualified for the flight in accordance with existing regulations. He had accumulated over 10000hours of flight experience, 131hours of which was on the Dash8. The captain attended Dash8 training at Flight Safety Canada and received a Dash8 rating on 14April2005 after 36hours of simulator and 2hours of flight training. The captain had five days of time off before the occurrence flight, and considered himself to be rested before the flight. The first officer held a valid ATPL and was qualified for the flight in accordance with existing regulations. He had 112hours of flight experience on this type of aircraft. He attended Dash8 training with the captain, and received a Dash8 rating on 14April2005 after 28hours of simulator and 1hour of flight training. The first officer had two days of time off before the occurrence flight, and considered himself to be rested before the flight. Company pilots receive training in aircraft "upsets" and unusual attitude recovery during initial and recurrent training. On 12March2005, both the captain and first officer completed a surface contamination and airborne icing course. Aircraft Information The aircraft's weight and centre of gravity were within the prescribed limits, and the aircraft was certified for flight in known icing conditions. There were no aircraft technical malfunctions that contributed to this occurrence. The aircraft's de-ice system removes ice accumulations from the leading edges of the wings, the horizontal and vertical stabilizers, and the inlet lip of the engine nacelles by alternately inflating and deflating pneumatic boots. The anti-ice systems use electrical heating elements to prevent ice formation on the left and right pitot probes and static ports, the left and right stall warning transducer vanes, the left and right engine compressor intake flanges, the elevator horns, the leading edge of the propeller blades, and the windshield. All ice protection systems were serviceable. The aircraft is configured with a SperrySPZ-8000 digital automatic flight control system (AFCS). A single flight guidance controller (see Figure1) is used to select modes of operation and engage/disengage the autopilot. Most of the controls on the AFCS controller are alternate-action push-button (pushON, pushOFF). There are two vertical modes available. The IAS button captures and holds the aircraft's indicated airspeed at the time of selection. The vertical speed (VS) button captures and holds the aircraft's VSat the time of selection. When VSmode is engaged, the airspeed is not controlled by theAFCS. If a change is required when either of these modes are engaged, the pitch wheel is used to dial in the new reference. The aircraft's trim is then adjusted automatically nose up or nose down to the new value. An ID-802 advisory display in front of both pilots shows, inter alia, the selected vertical mode (either IASorVS), its value (in KIASor hundreds of feet per minute), and the outside air temperature (in degrees Celsius). Flight Safety Canada standard operating procedures (SOPs) for the climb phase, page10.4, state the following: To help guard against inadvertent selection of VS mode and subsequent low airspeed, Flight Safety Canada SOPs require a verbal challenge and response when an AFCSmode is engaged. For example, upon engaging the IASmode, the pilot flying calls out, "SetIAS," along with the captured airspeed. The pilot monitoring confirms the selection of IAS and reads back the captured IASvalue. At the time of the occurrence, Provincial Airlines Limited SOPs for the climb phase did not warn against the use of VS mode during the climb; however, it was common knowledge that VS was not to be used during the climb. There was no requirement in the Provincial Airlines Limited SOPs for a verbal challenge and response between crew members for an AFCS mode engagement. The aircraft was equipped with a Fairchild F800FDR (part number 17M800-261, serial number03422). The recorder contained an expanded parameter data set that added pitch, roll, and yaw flight control position sensing, roll and pitch control disconnects, and propeller ground range beta. The FDR was shipped to the Transportation Safety Board of Canada (TSB) Engineering Laboratory for download and analysis of the data. The data quality on the tape was poor. Half of the data contained numerous drop-outs and data spikes. However, the data trend was still visible, and the data were vital to the investigation. In order to clean up the data, all the bad data from each parameter were deleted manually, allowing a linear interpolation between data points. The result was a clean trace on the plot, but a reduced resolution and frequency for each parameter. Only the data from the incident flight were corrected, due to the time required to clean up the data. The cockpit voice recorder did not have useful information relevant to the occurrence because it had been overwritten during post-occurrence flying. Additional Information Flight in Icing Conditions According to the aircraft flight manual (AFM) and company SOPs, icing conditions exist when the aircraft is flying in visible moisture in temperatures below 5C. When operating in icing conditions, or when ice is detected, engine intake bypass doors must be open and engine ignition switches are to be set at manual. The AFMrequires that the airframe de-ice be selected to slow or fast on initial detection of ice. The aircraft is not equipped with an electronic ice detector. Crews detect ice visually, looking for evidence of ice accumulation on the wing leading edges and on an ice probe located in front of the cockpit window. The Flight Safety Canada SOPs advise that, even if ice is not detected visually, ice may be present on portions of the aircraft that cannot be seen. Monitoring Errors A past study2 has noted that, when flight crews are monitoring automated systems, they may not be aware of the aircraft's energy state, particularly when approaching or trending toward a low-energy state. Monitoring errors occur more frequently if a pilot is engaged in some other manual task or where automation is highly reliable, leading to automation complacency. Monitoring errors can take place during both high and low workload situations. While these types of errors are common, they are also easily detected, and the industry has recognized the steps that may most effectively mitigate the risk of these types of errors occurring. For example, given the ease with which incorrect automation modes may be selected, the importance of crew procedures to cross-check mode selections has been highlighted: Where systems are in place that are vulnerable to mode errors, procedural constraints (such as limiting the number of modes routinely used and requiring that mode changes be announced and confirmed by both pilots) can become effective tools to mitigate the potential for mode errors to arise.3 Pneumatic De-ice Boot Operating Procedures For years, it was believed that, if the pneumatic boots were activated too soon, the ice would not break off, and the boots would subsequently inflate and deflate beneath an expanding ice bridge. However, research has shown that ice bridging does not occur with modern pneumatic boots, and the Dash8 flight manual reflects the current practice of "early and often" for pneumatic de-icing equipment. Transport Canada's (TC) Commercial and Business Aviation Advisory Circular (CBAAC) No.0130R, issued on 15June1999, included information resulting from investigations into accidents in which airborne icing was determined to be a contributing factor. Air operators were informed that they must amend their training programs to include the new information prior to 01 October1999. The section titled "10- Operational Use of Pneumatic De-Icing Boots"4 states: Provincial Airlines Limited manuals and programs were in compliance with TC's CBAAC No.0130R. However, in the course of the investigation, it became apparent that a sizeable proportion of Dash8 pilots may still hold to the traditional practice of waiting for ice to accumulate prior to activating the pneumatic boots, despite contrary instructions in guidance material. When contacted, Flight Safety Canada training personnel confirmed this, and estimated that 50percent of pilots, both Canadian and international, who attend their training sessions still hold to old practices despite directions to select de-icing equipment immediately upon entering icing conditions. Throughout the aviation industry, flight crews receive mandatory ground training in airborne icing. However, the ability to train in a simulator for flight in actual icing conditions is limited. The changes to stall characteristics with ice accumulation typically are not duplicated in simulator training, including the increase in stall speed and the onset of the stall before the activation of the artificial stall warning. Also, it is difficult to account for changes to normal stall symptoms, such as buffet or an increased tendency for a wing drop. While it is possible to teach these characteristics and the associated recovery techniques during classroom ground training, without the benefit of having experienced these stall symptoms, pilots can be ill-prepared to recognize and recover from contaminated-wing stall symptoms. The following Engineering Laboratory report was completed: LP 050/05 - Flight Data Recorder Analysis This report is available from the TSB upon request.